正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

A self-evolving agent + a 428B model + 3 Macs = ? Your own AI lab. We ran MiniMax (official) M3 locally with Parallax, right on our desk. Then GenericAgent took over to create a 5-stock portfolio and write it to disk. No cloud. No API bills. Nothing left... show more

Gradient

698,533 subscribers

153,169 次观看 • 2 个月前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

34 条评论

rw ./ 的头像

rw ./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai

Ivan Fioravanti ᯅ 的头像

Ivan Fioravanti ᯅ2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Wait… what is this magic? 👀

0xBagsAI_Sol./🌐 的头像

0xBagsAI_Sol./🌐2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Woww. Zero to hero. Local AI feel real!

Z./ 的头像

Z./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai zero cloud or API costs. Agent doing real work by itself. This is it ./

Hexx ./ 的头像

Hexx ./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai pump it

Pascal Wandera ./ 的头像

Pascal Wandera ./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai No cloud. No API bills. Interesting.

ANIEMA 🇳🇬./ 的头像

ANIEMA 🇳🇬./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Everyday is a progress with @Gradient_HQ Keep building Team. ./💚

Deez 的头像

Deez1 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Scammmm alert 🚨

AW3D 的头像

AW3D2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai cool stuff!

Jack sparrow 的头像

Jack sparrow2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Hey #Gradient long time no see 🙈

Phil 的头像

Phil2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Looks great! The future looks bright (i cannot afford one of them, even with my 7to5.)

vishal agarwal Spark light 的头像

vishal agarwal Spark light2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Good

Dhruv ฅᨐฅ ./ 的头像

Dhruv ฅᨐฅ ./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai This is the kings game we play

jmike.var 的头像

jmike.var2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai > no cloud. no API bills. team cooking

Logan Allen | Pumpkin.xyz Vaults 的头像

Logan Allen | Pumpkin.xyz Vaults2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai But running a 428B model on 3 Macs risks thermal throttling.

Ranzi$TOR 的头像

Ranzi$TOR2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai ./

Salah talib 的头像

Salah talib2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai innovation at its peak

K@rlitos 的头像

K@rlitos2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Woww...nice

DEATHLESS./ 的头像

DEATHLESS./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai

Qiu(❖,❖)./ 的头像

Qiu(❖,❖)./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai

realsir ./ 的头像

realsir ./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai build open ./

Muneeb Hussain Modi 的头像

Muneeb Hussain Modi2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Whatttt?? - Dario Amodei

linus cat 的头像

linus cat2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Can't take it anymore, why did you choose minimimax instead of deepseek

Rilwan M. B Sadiq ./🚢 的头像

Rilwan M. B Sadiq ./🚢2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai This is amazing. Keep building.

Jason傑森 🇭🇰 | 🛠️ 的头像

Jason傑森 🇭🇰 | 🛠️2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai 还挺酷

S T R O M / 的头像

S T R O M /2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai ./

Midnight ./ 的头像

Midnight ./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Omo

kozh ./ 的头像

kozh ./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai

Mayner (❖,❖) 的头像

Mayner (❖,❖)2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Wow Amazing Gradient 😎

Roman lemod ./ (✱,✱) zetarium 🟣 的头像

Roman lemod ./ (✱,✱) zetarium 🟣2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai superGrad./

Yehezkiel Kevin Fergiawan ./🌐 的头像

Yehezkiel Kevin Fergiawan ./🌐1 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Magic

Pascal2_22./ 的头像

Pascal2_22./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Sovereign AI is inevitable.

𝓁ℯ𝒾𝓁𝒶 ./ 的头像

𝓁ℯ𝒾𝓁𝒶 ./2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Hot

ℤãç𝕜𝕪 ™ ♰ 👀 的头像

ℤãç𝕜𝕪 ™ ♰ 👀2 个月前

@MiniMax_AI @tryParallax @GA_agent_ai Can I add it as an agent or model in vs code

相关视频

Create your own local and free AI agent using an open source model You can combine: - IBM's Granite 3.3 8B AI model - LM Studio to run it on a laptop - Smolagents to build your agent Small AI models are now powerful enough to run an autonomous agent. Thanks to IBM France for sponsoring this content!

Create your own local and free AI agent using an open source model You can combine: - IBM's Granite 3.3 8B AI model - LM Studio to run it on a laptop - Smolagents to build your agent Small AI models are now powerful enough to run an autonomous agent. Thanks to IBM France for sponsoring this content!

Paul Couvert

18,890 次观看 • 1 年前

Run your coding agent on a local model, with QVAC. OpenCode now plugs into a model on your own machine through our official plugin. That changes two things. Privacy by default. Your repo, your prompts, and every command the agent runs stay on your device. No third-party model ever sees your code. For an agent that reads your codebase and runs commands, that is the point, not a nice-to-have. New workflows. With no meter and no rate limit, the agent does not have to stop. Let it loop on a task, refactor a whole repo overnight, or run a watch-build-fix loop that keeps going. Setup takes a few minutes: one line in opencode.json and OpenCode runs on a local QVAC model. In the video demo, it handles three real developer tasks back to back, fully offline.

Run your coding agent on a local model, with QVAC. OpenCode now plugs into a model on your own machine through our official plugin. That changes two things. Privacy by default. Your repo, your prompts, and every command the agent runs stay on your device. No third-party model ever sees your code. For an agent that reads your codebase and runs commands, that is the point, not a nice-to-have. New workflows. With no meter and no rate limit, the agent does not have to stop. Let it loop on a task, refactor a whole repo overnight, or run a watch-build-fix loop that keeps going. Setup takes a few minutes: one line in opencode.json and OpenCode runs on a local QVAC model. In the video demo, it handles three real developer tasks back to back, fully offline.

QVAC

11,382,936 次观看 • 3 个月前

Run Cline on Local AI models with Atomic Chat! Cline is a coding agent trusted by 8M+ developers. Write, refactor, ship code securely on your own hardware with local models powered by atomic.chat — no cloud, private, free and open-source

Run Cline on Local AI models with Atomic Chat! Cline is a coding agent trusted by 8M+ developers. Write, refactor, ship code securely on your own hardware with local models powered by atomic.chat — no cloud, private, free and open-source

atomic.chat

44,433 次观看 • 3 个月前

Comfy MCP is now local and open-source! The #1 ask after Comfy Cloud MCP shipped in June. Your agent sees your actual install. So every node, custom nodes included, and every model on disk. It fetches files, starts your instance, gets a workflow to the point where it runs. It is now the easiest way to help with your local Minimax H3 workflows! Cloud MCP still does everything it did. Tell your agent where a job goes, or let it decide.

Comfy MCP is now local and open-source! The #1 ask after Comfy Cloud MCP shipped in June. Your agent sees your actual install. So every node, custom nodes included, and every model on disk. It fetches files, starts your instance, gets a workflow to the point where it runs. It is now the easiest way to help with your local Minimax H3 workflows! Cloud MCP still does everything it did. Tell your agent where a job goes, or let it decide.

ComfyUI

182,565 次观看 • 1 个月前

THIS FREE OPEN-SOURCE REPO GIVES YOU A SELF-HOSTED PERSONAL AI AGENT WITH MEMORY, WEB SEARCH, SUBAGENTS AND SCHEDULED AUTOMATIONS. NANOBOT RUNS WITH YOUR OWN API OR LOCAL MODEL AND CAN EVEN USE YOUR TERMINAL TO GET REAL WORK DONE.

THIS FREE OPEN-SOURCE REPO GIVES YOU A SELF-HOSTED PERSONAL AI AGENT WITH MEMORY, WEB SEARCH, SUBAGENTS AND SCHEDULED AUTOMATIONS. NANOBOT RUNS WITH YOUR OWN API OR LOCAL MODEL AND CAN EVEN USE YOUR TERMINAL TO GET REAL WORK DONE.

0xMarioNawfal

41,842 次观看 • 10 天前

A 100% open-source alternative to n8n! Sim is a drag-and-drop UI for creating powerful AI agent workflows: - Runs locally on your machine - Works with local LLMs I built a stock market research agent & connected it to Telegram in minutes. Here's a step-by-step guide:

A 100% open-source alternative to n8n! Sim is a drag-and-drop UI for creating powerful AI agent workflows: - Runs locally on your machine - Works with local LLMs I built a stock market research agent & connected it to Telegram in minutes. Here's a step-by-step guide:

Akshay 🚀

176,625 次观看 • 9 个月前

Spent the week with Wolffish an AI agent that runs entirely on my own laptop. No cloud, no server, nothing leaves my machine. Set it up in about 5 minutes, plugged in my own model key, and told it to be my morning news desk. It scanned the web, cross-checked every story across outlets, and handed me one clean PDF. Local-first and open source. This is the direction.

Spent the week with Wolffish an AI agent that runs entirely on my own laptop. No cloud, no server, nothing leaves my machine. Set it up in about 5 minutes, plugged in my own model key, and told it to be my morning news desk. It scanned the web, cross-checked every story across outlets, and handed me one clean PDF. Local-first and open source. This is the direction.

Nawi

89,262 次观看 • 3 个月前

$META Manus is moving its AI agent from the cloud to your desktop with the new Manus Desktop app. “My Computer” puts a fully local AI agent directly on your machine pushing AI toward edge-native workflows.

$META Manus is moving its AI agent from the cloud to your desktop with the new Manus Desktop app. “My Computer” puts a fully local AI agent directly on your machine pushing AI toward edge-native workflows.

Shay Boloor

65,771 次观看 • 6 个月前

GEMMA 4 IS STARTING TO KILL A LOT OF AI SUBSCRIPTIONS. Local inference, multimodal support, strong coding, and no API bill is making “just run it on your own machine” feel way more real.

GEMMA 4 IS STARTING TO KILL A LOT OF AI SUBSCRIPTIONS. Local inference, multimodal support, strong coding, and no API bill is making “just run it on your own machine” feel way more real.

0xMarioNawfal

65,679 次观看 • 5 个月前

Here is a new example of a concreate real-world use case of local AI: Invoice data. 55 supplier invoices, PDFs and photos, dropped in as one folder. About two seconds each, read on the laptop. It uses a local large language model plus a vision model to be able to cover all types of documents. No upload, no API key, 100% free & local. Write-up:

Here is a new example of a concreate real-world use case of local AI: Invoice data. 55 supplier invoices, PDFs and photos, dropped in as one folder. About two seconds each, read on the laptop. It uses a local large language model plus a vision model to be able to cover all types of documents. No upload, no API key, 100% free & local. Write-up:

QVAC

512,787 次观看 • 28 天前

Qwen 3.6 35B running at over 100 tokens per second on my $5,399 MacBook Pro M5 Max. This is the best local AI model I have ever run. 128GB of unified memory. No cloud. No API costs. No rate limits. Just raw local inference at speeds I didn't think were possible on a laptop. This model is more intelligent than GPT 5 on benchmarks. Running locally. On a MacBook. For free after the hardware cost. I said local AI would never compete with frontier. I'm starting to rethink that. The gap is closing faster than anyone expected.

Qwen 3.6 35B running at over 100 tokens per second on my $5,399 MacBook Pro M5 Max. This is the best local AI model I have ever run. 128GB of unified memory. No cloud. No API costs. No rate limits. Just raw local inference at speeds I didn't think were possible on a laptop. This model is more intelligent than GPT 5 on benchmarks. Running locally. On a MacBook. For free after the hardware cost. I said local AI would never compete with frontier. I'm starting to rethink that. The gap is closing faster than anyone expected.

BridgeMind

62,800 次观看 • 3 个月前

Introducing Docker 🐳 AI Models for local + private LLMs in VSCode 🤩 You can now use Docker as a local model provider in VSCode, JetBrains, Cursor, and soon Visual Studio Enterprise. With Docker Model Runner (v4.40+), run AI models locally on your machine... no data sharing, no cloud dependency 👏 Just update Docker, open CodeGPT, select the model, click "Download", and enjoy 👉 Here’s how to update Docker:

Introducing Docker 🐳 AI Models for local + private LLMs in VSCode 🤩 You can now use Docker as a local model provider in VSCode, JetBrains, Cursor, and soon Visual Studio Enterprise. With Docker Model Runner (v4.40+), run AI models locally on your machine... no data sharing, no cloud dependency 👏 Just update Docker, open CodeGPT, select the model, click "Download", and enjoy 👉 Here’s how to update Docker:

Daniel San

65,228 次观看 • 1 年前

NVIDIA is giving away free access to 140+ AI models for a full year > most people building AI agents are paying $50-200/month for API access NVIDIA just made that argument irrelevant models you get: GLM 5.2, MiniMax M3, Nemotron-3-Ultra-550B-A55B, Kimi K2.7, and 130+ more setup: > step 1 - get your free key > go to > register -> bind phone -> copy API key > step 2 - add to Hermes agent > open Settings -> Model Provider -> Custom base_url = " api_key = "nvapi-xxxxxxxxxxxxxxxxxxxx" > step 3 - pick a model model = "z-ai/glm-5.2" model = "minimaxai/minimax-m3" model = "nvidia/nemotron-3-ultra-550b-a55b" model = "moonshot-ai/kimi-k2.7" > Hermes already has NVIDIA set as default base_url > paste the key and you're running instantly > works the same in Cursor and OpenCode > cost: $0 > limit: 40 req/min > expires: 1 year while everyone is paying for API access, this is sitting there for free with this many models in one place, it is also the easiest way to test which one actually fits your agent before you commit to a paid provider

NVIDIA is giving away free access to 140+ AI models for a full year > most people building AI agents are paying $50-200/month for API access NVIDIA just made that argument irrelevant models you get: GLM 5.2, MiniMax M3, Nemotron-3-Ultra-550B-A55B, Kimi K2.7, and 130+ more setup: > step 1 - get your free key > go to > register -> bind phone -> copy API key > step 2 - add to Hermes agent > open Settings -> Model Provider -> Custom base_url = " api_key = "nvapi-xxxxxxxxxxxxxxxxxxxx" > step 3 - pick a model model = "z-ai/glm-5.2" model = "minimaxai/minimax-m3" model = "nvidia/nemotron-3-ultra-550b-a55b" model = "moonshot-ai/kimi-k2.7" > Hermes already has NVIDIA set as default base_url > paste the key and you're running instantly > works the same in Cursor and OpenCode > cost: $0 > limit: 40 req/min > expires: 1 year while everyone is paying for API access, this is sitting there for free with this many models in one place, it is also the easiest way to test which one actually fits your agent before you commit to a paid provider

Mr. Buzzoni

121,850 次观看 • 2 个月前

THEY RAN A LOCAL AI MODEL ON A RASPBERRY PI THAT COSTS LESS THAN $300 no cloud no subscription no data center device smaller than your wallet output on a 27 year old iMac I tested local ai this week on a macbook and a phone the hardware keeps getting cheaper and the models keep getting smarter bookmark this if you’re still paying $20 a month for something your $300 raspberry pi can do

THEY RAN A LOCAL AI MODEL ON A RASPBERRY PI THAT COSTS LESS THAN $300 no cloud no subscription no data center device smaller than your wallet output on a 27 year old iMac I tested local ai this week on a macbook and a phone the hardware keeps getting cheaper and the models keep getting smarter bookmark this if you’re still paying $20 a month for something your $300 raspberry pi can do

leopardracer

73,534 次观看 • 4 个月前

HERMES AGENT VS OPENCLAW. a local ai onboarding flow test. a 3.9gb bonsai served on localhost, both agents upstream and latest, i point each one at the endpoint and watch which one even finds it. > hermes opens a provider menu, thirty plus options, local servers sitting right there next to the cloud ones, i hand it 127.0.0.1:8899, it verifies the endpoint, one model visible, auto-detects the model by name, bonsai-27b-q1_0, reads the context length straight off the server, saves it, and starts reasoning and firing real tool calls on my local model. no key. no friction. > openclaw has no menu. it goes hunting for a codex login, an openai key, finds none because there are none, prints no models available three times, defaults to openai/gpt-5.5, a cloud model it cannot reach, and dead ends on run auth login --provider openai. read that back. it asked me for an openai key. to run a model already running on my own machine. it never once looked at localhost. to be fair, openclaw can run local if you hand wire endpoint yourself. what it will not do is find the model already sitting on your box. hermes agent found it in one line. now the part i owe you. the auto-detect that just won, the model name read, the .gguf strip, the context length probe off the server, that is my code, it is in hermes agent main right now, authorship preserved, #2051 and #4218. the wizard fix that stops an agent from silently routing you to someone else's creds, the exact trap openclaw still falls into, mine too, #4210. i contribute to hermes agent, i told you that going in. one agent is built to talk to whatever you are running, the other is built to talk to a cloud api, so one found my model and ran it and the other asked me to log into openai. onboarding flow of both, mapped, below.

HERMES AGENT VS OPENCLAW. a local ai onboarding flow test. a 3.9gb bonsai served on localhost, both agents upstream and latest, i point each one at the endpoint and watch which one even finds it. > hermes opens a provider menu, thirty plus options, local servers sitting right there next to the cloud ones, i hand it 127.0.0.1:8899, it verifies the endpoint, one model visible, auto-detects the model by name, bonsai-27b-q1_0, reads the context length straight off the server, saves it, and starts reasoning and firing real tool calls on my local model. no key. no friction. > openclaw has no menu. it goes hunting for a codex login, an openai key, finds none because there are none, prints no models available three times, defaults to openai/gpt-5.5, a cloud model it cannot reach, and dead ends on run auth login --provider openai. read that back. it asked me for an openai key. to run a model already running on my own machine. it never once looked at localhost. to be fair, openclaw can run local if you hand wire endpoint yourself. what it will not do is find the model already sitting on your box. hermes agent found it in one line. now the part i owe you. the auto-detect that just won, the model name read, the .gguf strip, the context length probe off the server, that is my code, it is in hermes agent main right now, authorship preserved, #2051 and #4218. the wizard fix that stops an agent from silently routing you to someone else's creds, the exact trap openclaw still falls into, mine too, #4210. i contribute to hermes agent, i told you that going in. one agent is built to talk to whatever you are running, the other is built to talk to a cloud api, so one found my model and ran it and the other asked me to log into openai. onboarding flow of both, mapped, below.

Sudo su

23,816 次观看 • 2 个月前

Monad is joining us on our mission to kill subscriptions. Your agent can now call Apollo, ElevenLabs, and MiniMax—no account, no API key—and pay per call in USDC on the Monad network. No credits to buy upfront. No subscriptions to manage. Nothing expires. Your agent isn’t a seat. It shouldn’t have to sign up for one. Made possible by Monad and Monid

Monad is joining us on our mission to kill subscriptions. Your agent can now call Apollo, ElevenLabs, and MiniMax—no account, no API key—and pay per call in USDC on the Monad network. No credits to buy upfront. No subscriptions to manage. Nothing expires. Your agent isn’t a seat. It shouldn’t have to sign up for one. Made possible by Monad and Monid

Shengkun Ye

14,545 次观看 • 14 天前

small local model that falls apart in bloated agents like openclaw just runs like a wild horse in hermes agent. and that's not even my line, someone else called it that, i've just been quietly pointing people at this harness for months because it held up on everything i threw at it, 3b models all the way to one trillion params. watch this happen on my own machine. i pointed hermes agent at a local http endpoint, gemma 4 12b on my 3090 llama.cpp server, and it auto-detected the model and started working immediately. no config wrestling, no broken tool calls, no babysitting the output format, i typed in a url and it just went. the whole clip is exactly that, start to finish, no errors, no retries, butter smooth. and the tool calling, the one thing that quietly breaks most local setups, works here like it's nothing. it's not the model that's flaky, it's the harness around it. hermes agent is the first agent i've run that actually gets that right. one url, one local model on one card, and it runs like a wild horse.

small local model that falls apart in bloated agents like openclaw just runs like a wild horse in hermes agent. and that's not even my line, someone else called it that, i've just been quietly pointing people at this harness for months because it held up on everything i threw at it, 3b models all the way to one trillion params. watch this happen on my own machine. i pointed hermes agent at a local http endpoint, gemma 4 12b on my 3090 llama.cpp server, and it auto-detected the model and started working immediately. no config wrestling, no broken tool calls, no babysitting the output format, i typed in a url and it just went. the whole clip is exactly that, start to finish, no errors, no retries, butter smooth. and the tool calling, the one thing that quietly breaks most local setups, works here like it's nothing. it's not the model that's flaky, it's the harness around it. hermes agent is the first agent i've run that actually gets that right. one url, one local model on one card, and it runs like a wild horse.

Sudo su

27,339 次观看 • 3 个月前

OpenClaw + Qwen 3.5 just unlocked “Claude-level” local agents. No API bills, no cloud, no data leaving your machine. Here’s the setup flow. → Install the Ollama pre-release (needed for the newest Qwen build). → Pull Qwen 3.5 locally (it’s a big download). → Launch OpenClaw with the Qwen model. → Run one real workflow: publish, message, schedule. If you can run it locally, you own the whole stack. Save this video, you’ll set up a free private agent. Want the SOP? DM me. 💬

OpenClaw + Qwen 3.5 just unlocked “Claude-level” local agents. No API bills, no cloud, no data leaving your machine. Here’s the setup flow. → Install the Ollama pre-release (needed for the newest Qwen build). → Pull Qwen 3.5 locally (it’s a big download). → Launch OpenClaw with the Qwen model. → Run one real workflow: publish, message, schedule. If you can run it locally, you own the whole stack. Save this video, you’ll set up a free private agent. Want the SOP? DM me. 💬

Julian Goldie SEO

163,670 次观看 • 6 个月前

🎉 Congrats to MiniMax (official) on releasing MiniMax M3! Frontier coding and agentic capabilities, native image and video input, computer use, and a 1M-token context window, all in a single open model. At the heart of M3 is MSA, a new sparse attention architecture: instead of attending densely over the full KV cache, each query scores 128-token KV blocks and runs attention only over the top blocks. That is what makes 1M-token context practical to serve. M3 runs in vLLM with day-0 support, verified on NVIDIA and AMD hardware: ✨ MSA sparse attention with dedicated prefill and decode kernels ✨ 1M-token context serving with prefix caching and chunked prefill ✨ BF16 and MXFP8 checkpoints, with MoE backends for both Hopper and Blackwell ✨ Native multimodal input (image + video) ✨ Tool calling, reasoning parsing, and thinking-mode control for agent workloads Day-0 support like this is a true team effort. Grateful to the teams at MiniMax (official), NVIDIA AI, AI at AMD, and Inferact, and to the vLLM community for making it happen. 🙏 Deep dive into the implementation, kernel work, and deployment recipes: 🔗

🎉 Congrats to MiniMax (official) on releasing MiniMax M3! Frontier coding and agentic capabilities, native image and video input, computer use, and a 1M-token context window, all in a single open model. At the heart of M3 is MSA, a new sparse attention architecture: instead of attending densely over the full KV cache, each query scores 128-token KV blocks and runs attention only over the top blocks. That is what makes 1M-token context practical to serve. M3 runs in vLLM with day-0 support, verified on NVIDIA and AMD hardware: ✨ MSA sparse attention with dedicated prefill and decode kernels ✨ 1M-token context serving with prefix caching and chunked prefill ✨ BF16 and MXFP8 checkpoints, with MoE backends for both Hopper and Blackwell ✨ Native multimodal input (image + video) ✨ Tool calling, reasoning parsing, and thinking-mode control for agent workloads Day-0 support like this is a true team effort. Grateful to the teams at MiniMax (official), NVIDIA AI, AI at AMD, and Inferact, and to the vLLM community for making it happen. 🙏 Deep dive into the implementation, kernel work, and deployment recipes: 🔗

vLLM

40,306 次观看 • 3 个月前

AI at Meta's Muse Glimmer (30B dense, open-weights) launches with SGLang day-0 support. We got ~230 tok/s on a single RTX 5090, with NVFP4 + DFlash on. It also works out of the box on NVIDIA AI Developer RTX Pro 6000, DGX Spark, and MLX for Mac. Speed and reliability have always been the blockers for a local agent. SGLang and Muse Glimmer make a local agent that's blazing fast, always on, and gets things done. Go build and run a local agent on your own machine!🚀

AI at Meta's Muse Glimmer (30B dense, open-weights) launches with SGLang day-0 support. We got ~230 tok/s on a single RTX 5090, with NVFP4 + DFlash on. It also works out of the box on NVIDIA AI Developer RTX Pro 6000, DGX Spark, and MLX for Mac. Speed and reliability have always been the blockers for a local agent. SGLang and Muse Glimmer make a local agent that's blazing fast, always on, and gets things done. Go build and run a local agent on your own machine!🚀

LMSYS Org

43,625 次观看 • 1 个月前